Answers are grounded in this paper and cite their sources.
In recent years, voice phishing has become increasingly sophisticated and has rendered the detection of financial frauds challenging. To address this, we propose a time-aware sliding window–based machine learning evaluation framework for financial fraud detection. This framework integrates preprocessing, feature selection, sliding window generation, model training, and evaluation, enabling continuous assessment under changing data distributions. Using the IBM Transactions for Anti-Money Laundering dataset, we compared random forest, LightGBM, XGBoost, and CatBoost on groups with lower and a higher illicit ratios. The results show that sliding window–based training and evaluation achieve relatively superior performance compared to fixed training and evaluation methods. These findings demonstrate that applying the sliding window technique is effective for enhancing the reliability and practicality of real-world financial fraud detection.
No takes yet. Share an insight, caveat, or question.
Lee et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: